[gringoDan]

It seems like the person who wrote this analysis assumed unchanging variable costs, rather than recognizing that they will decrease over time.

1) Where did they get a $5 marginal SG&A number? This seems high to me and no source is given.

2) There is no way that, at scale, delivery costs are as high as $5/order.

The IRS standard rate for mileage in 2018 is 54.5 cents/mile. (Maybe not the most accurate metric to use, but a decent rule of thumb.) So this analysis assumes that, at scale, for the average order a driver travels 9.17 miles. This routing would be incredibly inefficient - Amazon should "batch" deliveries so that in a 10-mile "loop" a driver can drop off 3+ orders.

Furthermore, I'm not sure if Amazon drivers are employees or 1099 workers. If they're independent contractors, the delivery costs are irrelevant to Amazon, since they're passed through to the drivers.

3) Naturally, drivers utilization rates will increase as demand increases.